Manufacturing and Asset Reliability Engineer. Tech scrappy. Opinions are mine. Except the bad ones.
Oil and Gas
Joined June 2025
- Tweets1K
- Following491
- Followers89
- Likes13.7K
Bandit retweeted
[greeting a bunch of 1800s industrial engineers who have traveled into my kitchen]
"Anyway, that's how modern packing in distillation columns works, you should really get onto refinery technology. Anyway, who wants some scones?"
"I-is that an entire roll of aluminum?"
Bandit retweeted
When the safety camp talks about handing control of AI over to governments, I understand and appreciate that most of them mean this in good faith. They see it as a way to give the public more control over AI and a bigger stake in everything that is about to happen.
The problem is that a significant percentage of the population now feels that they have never, in their entire lives, lived under a government that was not adversarial toward them to some degree. Left or right. Conservative or liberal. Parties. Platforms. It doesn't really matter anymore. So when you suggest handing over the most powerful technology that has ever existed to the government, any government, this group will not see it in the positive way I believe most of you intend.
Bandit retweeted
Today, we're releasing AutoCAD-Bench.
A benchmark for measuring whether AI models can complete precise AutoCAD tasks using computer-use alone.
GPT 5.6 sol dominates with 46%, solving many basic and intermediate tasks one-shot.
Full report below.
@markov__ai
Bandit retweeted
This argument by @deanwball is being badly misunderstood. It's OK to disagree with it, but first you have to actually understand what he's saying.
He's saying: releasing the weights for a frontier-level model is effectively dumping.
Dumping is when you sell a product at significantly below cost in order to corner market share. It's illegal. The reason: dumping results in short-term consumer surplus, but long-term it prevents the formation of a competitive market and discourages capex outside of the dumper. Standard Oil famously did this in order to consolidate the oil market before it was broken up.
So why is he claiming releasing the weights of a frontier level model is basically dumping? Isn't he just describing open source?
His argument: it's not financially sustainable to train a frontier model and release the weights. In the long run, you will not be able to internalize enough of the gains given the cost of training a frontier model, because neoclouds and other inference providers will be able to outcompete you at actually serving the model. It costs an astronomical amount of money to train frontier models, and if everyone else can serve them, you don't capture enough of the surplus to pay for the training and R&D.
It's not like normal open source when you build some software and then release it and sell services on top of it. The amount of capex required for frontier-level models is an order of magnitude higher than normal software, which is why doing this at frontier level is so economically irrational.
Right now the Hong Kong stock market is ebullient enough that Chinese AI companies are not getting punished for the fact that they're all deeply, deeply unprofitable. Releasing model weights is great marketing, intellectually appealing, and strikes fear into the hearts of their opponents. We can assume the status quo continues for a while because of the AI supercycle. But eventually the AI market will correct, the Hong Kong market will dump, and suddenly these Chinese labs won't be able to afford to training super expensive models without internalizing more of the gains.
But what if China, seeing that this strategy is successfully kneecapping the US lead (by discouraging further capex and lowering valuations), says no--don't stop. And so the Chinese government starts buying up the shares of these companies and demanding that they continue releasing frontier-level weights, profitable or not.
In that case, it becomes a genuine space race. For-profit companies cannot continue to compete on either side. US labs valuations fall, and the White House realizes that to keep their advantage in the AI race, they cannot rely on the free market to maintain their lead. They nationalize the labs and fund them off government subsidies.
Now you have government-controlled and distributed models on both sides. That's what Dean is calling the "dystopian hellscape."
The best analogy is drug development: if China were to sell American drugs back to us really cheaply, that would result in a large short-term consumer surplus. Cheap Viagra and Ozempic is obviously great. But in the long run, this would discourage investment in developing new drugs. That's the sense that Dean is saying it's long-term "decel."
Now, I happen to disagree with Dean. I think the consumer surplus of having frontier-level open weight models is huge, even at the current capabilities. I also think China is going to defect from this strategy soon (there's been reporting along these lines, that Beijing will stop allowing large models to be open-weight; I think there are other reasons for this aside from competition). I also suspect that nationalization of labs is inevitable as they take on more geopolitical and cyber capabilities.
But he's not wrong--releasing frontier-level weight models is weird.
The question of how long this market will remain profit-driven is a very coherent question to ask.
Bandit retweeted
They talking about gpt 5.6 like a white guy in the NFL. Not so naturally talented but he's a gym rat. Just won't quit. Always watching tape
Bandit retweeted
We are leaving the Old Code Age, the Paleocodic, the artisinal code era, where if you needed a novel program, you would commission a local codesmith or code guild to hand-craft a work of code for you, bespoke.
Bandit retweeted
A few random thoughts on the Fable 5/GPT-5.6 situation:
1. I see some people on the timeline blaming Anthropic for scaring the USG into restricting access to these models. Frankly, that is nonsense. The USG is restricting access to these models because of their *real and dangerous* cybersecurity capabilities. Anthropic is blameless here.
If I were in government relations at Anthropic, I would be telling the leadership to proactively inform the government that Mythos Preview is dangerous before it's publicly released. This is alpha and omega of government relations: if there is a potential serious issue with one of your products, make sure that your regulator hears about this from *you*, not some third party.
(This is also why it was a huge mistake for Anthropic to let Amazon break the news about the jailbreak to the USG instead of self-reporting after first hearing about this issue from Amazon.)
2. On the other hand, Anthropic needs to learn that, if Bessent calls you with a potential national security issue, the leadership's immediate reaction should be "we absolutely hear you and will cooperate with you to the fullest extent right this minute, but can we please have some meetings afterwards to clarify how we can help you better and explain our position?"
You can't just tell a regulator "no". You can't just tell a regulator that its concern is not serious, actually.
3. I've been incredibly impressed by the work being done by Tom Brown in getting this situation resolved. I think he and his team will get it done, and Fable 5 will be in our hands before too long.
I think Tom Brown's stock has risen a lot over the past few days.
4. More specifically on "before too long", the Executive Order requires the NSA to finalize the benchmark for doing "voluntary" testing of frontier models for cyber risk within 60 days of the date of the Executive Order - so, by early August. It seems that at least Anthropic and OpenAI (and maybe some of the other labs) are now working with the USG on developing both this benchmark and the "rules of the road" for testing and approving models in the future. I suspect that, once this benchmark and "rules of the road" are finalized, it will be announced that both GPT-5.6 and Fable 5 pass and can be released to the public.
5. I generally do not view this administration as being interested in holding frontier AI models back from the public unless the circumstances warrant it. This is particularly the case when keeping the models back for an extra few weeks is presumably giving the USG more time to use them defensively to squash bugs in its own systems.
When bio and chemical risks are on the table, this calculus may change.
6. I suspect that the industry will try to get the USG to publish a clear written process for mandatory (not "voluntary") testing of covered models, together with specific disapproval standards, a right of appeal, and some transparency around the process. I also suspect that the USG will push against this and want to keep the process in line with the Executive Order - i.e., lacking specifics regarding the process itself, public transparency, or concrete disapproval standards (probably other than for cyber in terms of model capabilities only).
This will doubtlessly be blamed on the administration, but I am actually quite sympathetic to their predicament. Imagine being told by the labs that AI is close to improving recursively, the risks currently include cyber, but in the future could include just about anything else (including unknown unknowns), and that even the labs themselves can't easily predict the risks or their timing. As the USG, you are going to want maximum flexibility; you are not going to want specific written standards or any limits on your power to yank a model at a moment's notice. Much has been written about this specific administration's lack of transparency and trigger-happiness throughout this process, but I suspect that most other administrations (Democratic or Republican) would act much the same.
7. The biggest outcome of these events that no one seems to be thinking about is the prospect of China waking up. Anthropic has engineers embedded at the NSA, helping it to conduct offensive cyber operations. Against whom do you think these operations are being conducted?
If I'm the CCP, having found myself on the receiving end of some of these attacks, and having now read reports that multiple U.S. labs seem to have these kinds of capabilities, I get significantly concerned. I stop listening to my country's labs telling me that they're just 4 months behind the U.S. I start seriously thinking about how to catch up to the frontier, as fast as humanly possible.
Bandit retweeted
I think I see what's going on here. Anthropic is negotiating with a regulator without realizing it.
When you submit something risky to a regulator, you have to (1) describe what the risk is, and (2) show how you're going to mitigate that risk. Importantly, you have to properly estimate the scope of the risk. If you scope the risk too broadly, then you place extra burden on your mitigation. If you scope too narrowly, then you run the risk directly. But you must commit to an estimation to the regulator.
What happened here is that Anthropic scoped too broadly. And when it was shown that their mitigation was insufficient for the stated risk scope, they responded by trying to narrow it. This is a huge red flag to any regulator (for good reasons), so regulatory action followed.
Anthropic: "this could be a cyber-weapon, but we have these guard-rails to prevent that, so it's safe"
Government: "we accept your plan"
Amazon: "we broke the guard-rails"
Anthropic: "yeah but this particular jailbreak won't make it a cyber-weapon so it's fine"
Government: "that's not what you committed to"
Anthropic: "we didn't commit to any plan, only to safety in general"
Government: "we can't trust you with that"
I’ve had a number of conversations with folks inside and outside government about the current situation with Anthropic, and here is what I believe to be true:
— As we know, Anthropic publicly released its Mythos class models earlier this week under the commercial name Fable.
— Fable is Mythos with guardrails. But if those guardrails fail, then you’ve exposed Mythos and its advanced cyber capabilities to people who shouldn’t have them. (Keep in mind that Anthropic itself widely promoted the idea that Mythos was a cyberweapon and needed to be regulated as such. They asked for government regulation of Mythos and championed the guardrails on Fable. If there is a vulnerability — big or small — it is Anthropic’s responsibility to patch.)
— A highly credible trusted partner of both Anthropic and the USG who was testing Fable came forward with a jailbreak of those guardrails. The Admin asked Dario to fix the jailbreak or de-deploy the model. Dario refused.
— In their blog post, Anthropic defended its decision by saying the jailbreak isn’t serious. That is not what the trusted partner and the USG believe; nor is that kind of minimizing language consistent with Anthropic’s brand as the AI safety company. It’s difficult to fathom how they could claim a jailbreak allowing operability of a cyber weapon could be defined as not “serious.”
— In the past, Anthropic has always said that safety must be top priority and taken super seriously. In this case, Anthropic prioritized the continued offering of the consumer model over safety.
— In reaction, the Admin issued the export control. The Admin did this reluctantly. It’s been very surprised that Anthropic hasn’t wanted to cooperate with a reasonable safety request (ie fixing the jailbreak issue). Anthropic’s reaction is very much at odds with their branding and ethos as a safe AI research community.
— The Admin’s hope now is that Anthropic remediates the safety issue, the export control is lifted, and Fable goes back into general release. The Admin wants all of this to happen as soon as possible. It is frankly bewildered that Anthropic hasn’t wanted to comply with safety requests that it previously said were its highest priority.
— Those trying to misdirect and tie this action to the prior DoW/Anthropic issues are wrong. The Admin values Anthropic’s technical capabilities and feels that this issue, while serious, should be easily resolved. The ball is in Anthropic’s court.
Bandit retweeted
i’d like more clarity on:
- what, precisely, amazon told the administration;
- what level of risk the alleged jailbreak actually presented;
- whether independent cybersecurity researchers agree with amazon’s assessment (and if amazon even thought this level of intervention was needed);
- who was involved in the decision-making process (DoW, Commerce);
- what Anthropic did between being informed of the risk and being hit with export controls
Anthropic wants you to believe this is a personality or political dispute. It's not.
Anthropic abandoned its safety guardrails the moment they became commercially inconvenient. It was so egregious their corporate partner--Amazon--felt compelled to blow the whistle.
Now they're deploying a team to DC to do technical cleanup. Anthropic doesn't care about safety, they care about control. This is dangerous and wrong.
Bandit retweeted
the benchmark-obsessed still haven’t absorbed the lessons of o1
think of those accounts throwing tomatoes at old kings and crowning new ones every week, living for the next single-number decisive victory as the elonbucks roll in
what if all that effort misses the point
Bandit retweeted
when sam talks about the risks of compute scarcity and why we have to “flood the market” to avoid them, this is the kind of centralising force he worries about
one man should not be able to control the flows based his own ideas of moral worth
this power must be undercut
Bandit retweeted
In Vending-Bench Arena (the multiplayer version of Vending-Bench with competition dynamics), GPT-5.5 actually beats Opus 4.7.
Opus 4.7 showed similar behavior to Opus 4.6: lying to suppliers and stiffing customers on refunds. GPT-5.5's tactics were clean, and it still won.
Bandit retweeted
Starting doubt the Mythos narrative. If they have a God model then why:
- Is Opus 4.7 so bad
- Is Claude Code so buggy
- Is Claude Desktop unfinished
- Is the API down all the time
Bandit retweeted
A lot of the concerns about the government's "all lawful use" language seem to stem from mistrust that government will follow the laws. At the same time, people believe that Anthropic took an important stand by insisting on contract language around their redlines.
We cannot have it both ways. We cannot say that the government cannot be trusted to interpret laws and contracts the right way, but also agree that Anthropic’s policy redlines, in a contract, would have been effective.
This is why our approach has been:
Let the democratic process decide on the legality and proper use question. The fact that people can even say that the gov has made mistakes in the past is the process in action. The fact that we are having this discussion on twitter is part of the process.
Create a reasonable contractual framework that guides expectations and the relationship, just as much if not more than the rules themselves.
And on top of this, have the ability to build the models the way we think is safe, along with cleared FDEs to do the real world work in partnership.